Struct BucketView

source

pub struct BucketView<'a> { /* private fields */ }

Expand description

A view into a metrics bucket. Sometimes also called a partial bucket. A view contains a subset of datapoints of the original bucket.

            Full Bucket
/-------------------------------\
[0, 1, 2, 3, 5, 5, 5, 10, 11, 11]
\----------------/\-------------/
      View 1          View 2

A view can be split again into multiple smaller views.

Implementations§

source §

impl<'a> BucketView<'a>

source

pub fn new(bucket: &'a Bucket) -> Self

Creates a new bucket view of a bucket.

The resulting view contains the entire bucket.

source

pub fn timestamp(&self) -> UnixTimestamp

Timestamp of the bucket.

pub fn width(&self) -> u64

Width of the bucket.

pub fn name(&self) -> &'a MetricName

Name of the bucket.

pub fn value(&self) -> BucketViewValue<'a>

Value of the bucket view.

source

pub fn ty(&self) -> MetricType

Type of the value of the bucket view.

source

pub fn tags(&self) -> &'a BTreeMap<String, String>

Name of the bucket.

pub fn tag(&self, name: &str) -> Option<&'a str>

Returns the value of the specified tag if it exists.

pub fn metadata(&self) -> BucketMetadata

Returns the metadata for this bucket.

The aggregation process of metadata is inheritly lossy, which means some metadata, for example the amount of merges, can not be accurately split or divided over multiple bucket views.

To compensate for this only a bucket view which contains the start of a bucket will yield this metadata, all other views created from the bucket return an identity value. Merging all metadata from non-overlapping bucket views must yield the same values as stored on the original bucket.

This causes some problems when operations on partial buckets are fallible, for example transmitting two bucket views in separate http requests. To deal with this Relay needs to prevent the splitting of buckets in the first place, by never not creating too large buckets via aggregation in the first place.

pub fn len(&self) -> usize

Number of raw datapoints in this view.

pub fn is_empty(&self) -> bool

Returns true if this bucket view contains no values.

source

pub fn select(self, range: Range<usize>) -> Option<Self>

Selects a sub-view of the current view.

Returns None when:

the passed range is not contained in the current view.
trying to split a counter or gauge bucket.

source

pub fn estimated_size(&self) -> usize

Estimates the number of bytes needed to serialize the bucket.

Note that this does not match the exact size of the serialized payload. Instead, the size is approximated through the number of contained values, assuming an average size of serialized values.

source

pub fn split( self, size: usize, max_size: Option<usize>, ) -> (Option<Self>, Option<Self>)

Calculates a split for this bucket if its estimated serialization size exceeds a threshold.

There are three possible return values:

(Some, None) if the bucket fits entirely into the size budget. There is no split.
(None, Some) if the size budget cannot even hold the bucket name and tags. There is no split, the entire bucket is moved.
(Some, Some) if the bucket fits partially. Remaining values are moved into a new bucket with all other information cloned.

This is an approximate function. The bucket is not actually serialized, but rather its footprint is estimated through the number of data points contained. See estimated_size for more information.